Go back
What relative point values of pieces do engines use?

What relative point values of pieces do engines use?

Only Chess

MB

Joined
06 May 13
Moves
0
Clock
30 Jul 14
Vote Up
Vote Down

Hello.

Humans use:
Pawn = 1
Knight = 3
Bishop = 3
Rook = 5
Queen = 9 or 9.5
Bishop pair = +0.5

But I would like to know what relative point values do engines use (preferably strong engines, like Stockfish, Houdini, Rybka, Komodo, ...).

If the point value that the engine assign to the pieces depends on the position of the pieces on the board (for example a White Knight on a1 is only worth 2.7 but on e5 it's worth 3.3), than I would like to know the average point values of the pieces.

Thanks in advance for your answers.

BigDogg
Secret RHP coder

on the payroll

Joined
26 Nov 04
Moves
155080
Clock
30 Jul 14

Originally posted by Marc Benford
Hello.

Humans use:...
Bishop pair = +0.5
That's a new one for me.

Ponderable
chemist

Linkenheim

Joined
22 Apr 05
Moves
669903
Clock
30 Jul 14
Vote Up
Vote Down

I think the basic algorithms use the human values. And then points are added adn subtracted for position, but I am just a layman on that.

KnightStalker47

Joined
12 Nov 06
Moves
74414
Clock
30 Jul 14
1 edit

The points system really doesn't tell the full story. In a world without endgames, I would give the rook 3.5-4 points. I mean you can sac a rook for a knight in the first 15 moves and you wont really feel it until the ending. Same could be said for pawns.

The bishop pair is an advantage in theory, they cover each others weaknesses. But I have never seen anyone win a game because of a bishop pair. It might be +0.1 or +0.2, but taking it to +0.5 is way too far IMO.

Paul Leggett
Chess Librarian

The Stacks

Joined
21 Aug 09
Moves
114051
Clock
31 Jul 14
1 edit
Vote Up
Vote Down

Originally posted by KnightStalker47
The points system really doesn't tell the full story. In a world without endgames, I would give the rook 3.5-4 points. I mean you can sac a rook for a knight in the first 15 moves and you wont really feel it until the ending. Same could be said for pawns.

The bishop pair is an advantage in theory, they cover each others weaknesses. But I ha ...[text shortened]... me because of a bishop pair. It might be +0.1 or +0.2, but taking it to +0.5 is way too far IMO.
According to research by GM John Emms, the bishop pair tends to win 60-62% of the time. That's more valuable than being white.

I'm definitely on board with your opinion on rooks. Really the bottom line is that all piece and pawn values are relative to the position on the board.

C
Cowboy From Hell

American West

Joined
19 Apr 10
Moves
55013
Clock
31 Jul 14

All piece point values are meaningless without considering their place on the board and interaction with other pieces. I do know one thing. I've noticed over the years players accessing the position (in OTB play) and looking at captured pieces. Pieces off the board have no value at all. An unstoppable pawn on the 7th rank has much more value than 1. A queen that is trapped and powerless can not have a value of 9.

D
Losing the Thread

Quarantined World

Joined
27 Oct 04
Moves
87415
Clock
31 Jul 14
Vote Up
Vote Down

Originally posted by Marc Benford
Hello.

Humans use:
Pawn = 1
Knight = 3
Bishop = 3
Rook = 5
Queen = 9 or 9.5
Bishop pair = +0.5

But I would like to know what relative point values do engines use (preferably strong engines, like Stockfish, Houdini, Rybka, Komodo, ...).

If the point value that the engine assign to the pieces depends on the position of the pieces on the boa ...[text shortened]... ould like to know the average point values of the pieces.

Thanks in advance for your answers.
Those are basically the values that programmers give to their engines, but with the same variation as for humans. What is possible, and I don't know if it has been done, is to vary the values and see how it affects the performance of the engine. They may well not have done as engines basically work by calculating deeply and making a rough assessment of the final position. It is the depth that they can calculate lines to which tends to matter, not how accurate the assessment of the final position is.

You can download the source code for Crafty and look in there to see which values it uses.

Cookies help us deliver our Services. By using our Services or clicking I agree, you agree to our use of cookies. Learn More.